agentleFS
awaikawaik/direct-mcp-ai-project/.agents/skills/workspace-project-manager/SKILL.md

workspace-project-manager

Skill20 starsChanged 3 days agoMITView on GitHub

Управлять Пространствами LidFly и project-first памятью Workspace: проекты, provider entities, кампании, документы, решения, настройки, задачи и AI-автозапуски. Использовать для любой записи памяти или управления проектом с точным workspace_project_id.

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 4 are distinctive, and 13 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 13

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use for projects, clients, memory, decisions, documents, settings, saved campaign context, tasks, reminders and scheduled AI checks.

    line 8
  2. Distinctive:

    Before writing audits, documents, decisions, snapshots, settings, provider/campaign links or provider-scoped tasks:

    line 19
  3. Very specific:

    If exact workspace_project_id is known, use it.

    line 21
  4. Very specific:

    If resolved, write with returned workspace_project_id.

    line 23
  5. Very specific:

    If ambiguous, show candidates and ask for exact workspace_project_id.

    line 24
  6. Very specific:

    If no project matches, offer workspace_create_project.

    line 25
  7. Very specific:

    Call top-level get_provider_context with the exact workspace_project_id before treating a missing provider scope as an error.

    line 32
  8. Very specific:

    If it returns provider_link_candidates, those scopes are verified but not executable inside the project yet.

    line 33
  9. Very specific:

    When the user asked to link the cabinet, choose the exact candidate and execute only its prepared next_action through call_write_tool.

    line 34
  10. Very specific:

    If several candidates are returned, ask the user which cabinet to link.

    line 35
  11. Distinctive:

    Read-only project members must not receive unlinked owner OAuth candidates.

    line 37
  12. Very specific:

    Never pass top-level meta-tools such as search_tools or get_tool_schema as tool_name.

    line 63
  13. Very specific:

    If can_delete=false, explain the returned blocker.

    line 71
  14. Very specific:

    If ready, show confirmation_message, the deletion counts, and retained activity-history count.

    line 72
  15. Very specific:

    Call workspace_delete_project through call_write_tool with the unchanged workspace_project_id, expected_project_name, and …

    line 73
  16. Distinctive:

    If the target changed, run the preflight again and request a new confirmation.

    line 74
  17. Very specific:

    If a future check must be shown to the owner, asks a question, or needs a new decision or confirmation, use workspace_add_tasks.

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Workspace Project Manager

Use for projects, clients, memory, decisions, documents, settings, saved campaign context, tasks, reminders and scheduled AI checks.

Model

  • Пространство is the user-facing memory container.
  • Workspace project is a business, project, direction, or agency client.
  • workspace_project_id is the canonical id for memory writes.
  • Provider identifiers are external entities, not project ids: client_login, client_id, account_id, counter_id, host_id, subdomain.

Project Scope

Before writing audits, documents, decisions, snapshots, settings, provider/campaign links or provider-scoped tasks:

  1. If exact workspace_project_id is known, use it.
  2. Otherwise call workspace_prepare_project_scope with all known selectors: project_name, provider, external_entity_key, external_campaign_id, client_login, vk_client_id, account_id, metrika_counter_id, lidfly_subdomain, host_id, campaign_name.
  3. If resolved, write with returned workspace_project_id.
  4. If ambiguous, show candidates and ask for exact workspace_project_id.
  5. If no project matches, offer workspace_create_project; never silently create "Основной проект".
  6. Project resolution selects a destination, not write permission. Write only on the user’s request or under a memory workflow authorized for this exact operation. Never capture another client’s data, style or decisions.

For a business brief or campaign history, read this guide.

Provider Links

  • Call top-level get_provider_context with the exact workspace_project_id before treating a missing provider scope as an error.
  • If it returns provider_link_candidates, those scopes are verified but not executable inside the project yet.
  • When the user asked to link the cabinet, choose the exact candidate and execute only its prepared next_action through call_write_tool. Keep its arguments unchanged; MCP confirmation is still required.
  • If several candidates are returned, ask the user which cabinet to link. Never invent client_login from the project/account name or external_entity_key.
  • A missing provider link is a normal confirmation-gated write workflow, not a support incident. Do not call support_prepare_report for it.
  • Read-only project members must not receive unlinked owner OAuth candidates; do not try to recover them through personal scope enumeration.

Tools To Prefer

Find Workspace tools with search_tools({ provider: "workspace", ... }); read each schema with get_tool_schema before its first call.

Read through call_tool:

  • workspace_list_projects
  • workspace_get_project
  • workspace_prepare_project_scope
  • workspace_prepare_project_deletion
  • workspace_get_settings
  • workspace_get_tasks
  • workspace_get_scheduled_ai_tasks

Write through call_write_tool:

  • workspace_create_project
  • workspace_delete_project
  • workspace_upsert_provider_entity
  • workspace_link_campaign
  • workspace_update_settings
  • workspace_add_tasks
  • workspace_schedule_ai_task

Never pass top-level meta-tools such as search_tools or get_tool_schema as tool_name.

Permanent Project Deletion

Permanent deletion is owner-only, irreversible, and never allowed in an AI autostart.

  1. Resolve and use the exact workspace_project_id; never select a deletion target from a similar name.
  2. Call read-only workspace_prepare_project_deletion.
  3. If can_delete=false, explain the returned blocker. Archive an active project only if the user asked; protected accounting history means the project must remain archived.
  4. If ready, show confirmation_message, the deletion counts, and retained activity-history count. Wait for an explicit textual confirmation.
  5. Call workspace_delete_project through call_write_tool with the unchanged workspace_project_id, expected_project_name, and expected_updated_at from that preflight.
  6. If the target changed, run the preflight again and request a new confirmation. After a transport-uncertain delete, reread the exact project before considering any retry.

Reminders vs AI Autostarts

  • workspace_add_tasks is a manual reminder: it stores a prompt and due date, but the due date only triggers email and never runs AI or provider tools.
  • If a future check must be shown to the owner, asks a question, or needs a new decision or confirmation, use workspace_add_tasks.
  • workspace_schedule_ai_task is an AI autostart: LidFly executes the saved plan automatically at the specified time without a new confirmation.
  • Objects, actions, values, and all conditional branches must be fully approved before using workspace_schedule_ai_task.

For workspace_schedule_ai_task:

  • allowed_tools must list real domain tools for the future run, not v3 meta-tools.
  • Include workspace_project_id for provider/campaign tasks.
  • For write future runs, include concrete target items and a confirmed plan.

Output

Summarize the selected project, created object type, whether it runs automatically, next user action and unconfirmed details.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.